Earnings volatility-variability over time in a worker's earnings-is interesting for its potential welfare consequences and as a labor market outcome. In the presence of liquidity constraints, households may not be able to smooth consumption when faced with volatile earnings. And rising earnings volatility may indicate that workers are changing jobs more often or that implicit contracts governing compensation within jobs are smoothing pay less, to name two possibilities. This dissertation shows that earnings volatility has increased in the U.S., and assesses, for a specific concept and measure of volatility, the reasons for the increase. After an introduction, Chapter 2 introduces my preferred volatility measure, the volatility of residual earnings after estimating life cycle earnings profiles. Previous literature covers an array of concepts and measures of volatility. I outline differences among these approaches and argue that the primary motivation for studying volatility-potential welfare losses-has implications for the specification of the life cycle profiles and measures of earnings volatility. Chapter 3 describes the data I use, a sample of male heads of household from the PSID. In Chapter 4, I show that earnings volatility has increased in the U.S., and that about 70 percent of the increase is explained by volatility of wages rather than hours worked. I describe how earnings volatility differs across groups, and show that it has increased among almost all groups. Finally, I consider whether these findings depend on some measurement choices. All measures show increasing earnings volatility in the U.S., but the amount of the increase and comparisons of volatility across groups are often sensitive to methodological choices. Chapter 5 addresses why earnings volatility has increased in the U.S. I develop a decomposition approach to attribute changes in economy-wide volatility to various factors. I create a demand shock index that measures workers' predicted exposure to labor demand shocks, using national changes in the occupation-industry distribution of hours worked. My major finding is that larger or more frequent labor demand shocks explain about half of the increase in economy-wide earnings volatility between 1975 and 2005.